Is This Time Different? The Slowdown in Health Care Spending

Abstract

Why have health care costs moderated in the last decade? Some have suggested that the Great Recession alone was the cause, but health expenditure growth in the depths of the recession was nearly identical to growth prior to the recession. Nor can the Affordable Care Act (ACA) take credit, since the slowdown began prior to its implementation. Instead, we identify three primary causes of the slowdown: the rise in high-deductible insurance plans, state-level efforts to control Medicaid costs, and a general slowdown in the diffusion of new technology, particularly for use by the Medicare population. A more difficult question is: Will this slowdown continue? On this question we are pessimistic, and not entirely because a similar (and temporary) slowdown occurred in the early 1990s. The primary determinant of long-term growth is the continued development of expensive technology, and there is little evidence of a permanent slowdown in that pipeline. Proton beam accelerators are on target to double between 2010 and 2014, while the market for heart-assist devices (costing more than $300,000 each) is projected to grow rapidly. Accountable care organizations (ACOs) and emboldened insurance companies may yet stifle health care cost growth, but our best estimate over the next two decades is that health care costs will grow at GDP plus 1.2 percent, a rate lower than previous estimates but still on track to cause serious fiscal pain for taxpayers and workers who bear the costs of higher premiums.

The United States has led the world in both the level and growth rate of spending on health care, with nearly 18 percent of U.S. GDP currently produced by the health sector. Between 1980 and 2008, U.S. health care spending grew by 7 percentage points of GDP, compared to an average of just 2.6 percentage points among all other OECD countries combined (Chandra and Skinner 2012). Yet recent time-series evidence suggests that health care cost growth in the United States is now moderating. David Cutler and Nikhil Sahni (2013) estimate an unexplained one-percent drop in health care spending growth, while Alexander Ryu and others (2013) find a large downturn in the growth in private insurance spending over the last several years.

Others are less sanguine. Joseph Antos (2013) recently testified before a Senate committee that “[t]he biggest single factor driving the recent slowdown is the economy,” implying that health care spending will pick up once the economy recovers (Kaiser Family Foundation 2013). Similarly, Charles Roehrig and Gene Steuerle have documented that health care expenditure growth has exhibited a remarkably stable pattern relative to GDP growth, implying convergence toward a steady state in which nearly one-third of GDP is devoted to health care (Steuerle 2013; Roehrig 2013). Still others suggest that analysts are wasting their time trying to distinguish between these two hypotheses, since one cannot predict long-term trends based on a few years of data (Fuchs 2013).

In this paper, we ask the question, Is the slowdown really different this time?¹ We first study trends in a variety of measures of U.S. health care, including personal health expenditures, total health spending, health care prices and quantities, and employment in the health care sector. While the different measurement approaches tell somewhat different stories about inflection points, nearly all of them point to a recent decline in health care spending, with the exception of one: the Current Employment Statistics (CES) shows little slowdown in health care employment growth relative to the size of the sector.²

Previous studies used time-series regressions to test whether current or lagged GDP growth affects current health care spending. On theoretical grounds, we question whether GDP growth alone should have a large impact on health care spending—short-run income effects for health care spending are modest at best (McClellan and Skinner 2006). Instead, we hypothesize that three factors have led to the decline in health care expenditure growth.

The first factor in the decline in cost growth has been the rapid diffusion of high-deductible health care plans, which, together with a continued decline in the fraction of people insured, has led to higher out-of-pocket prices for health care. In this new regime, households did scale back physician visits, particularly households in financial distress (Lusardi, Schneider, and Tufano 2010).

The second factor in the decline has been cuts in Medicaid benefits and reimbursement rates that were necessitated by shrinking state budgets, with the end result of nearly flat per capita real spending for the growing population of Medicaid enrollees. As in other OECD countries facing similar budgetary pressure (OECD 2013), state Medicaid officials in the United States restricted services and cut reimbursement rates, which in turn attenuated the utilization of specialists, many of whom no longer accepted Medicaid patients (Sack 2010).

The third factor stems from the use of medical technology for people under Medicare. As an entitlement program, utilization in Medicare is not subject to either budgetary cutbacks (as in Medicaid or in other OECD countries), nor to rising out-of-pocket costs (as in private insurance). Why then has Medicare spending moderated? As Levine and Buntin (2013) show, the Medicare slowdown growth could not be attributed to elderly people being in better health, nor could it arise from financial stress among Medicare enrollees. Instead, we present evidence that a third factor underlying the reduction in health care growth has been a slowdown in the diffusion of those new technologies that had accounted for much of the growth in expenditures during the previous decade. For example, angioplasty (the insertion of a balloon into a clogged cardiac artery) and, later, the use of stents (cylindrical wire devices designed to maintain blood flow in the heart’s arteries) grew at double-digit rates during the 1990s and early 2000s, but their use has since stagnated. Similarly, Cutler and Sahni (2013) document a decline in the introduction of new blockbuster drugs and the transformation of former brand drugs to generic status.

Will the slowdown in cost growth persist? Medicaid programs have been successful in cutting reimbursement rates and restricting benefits, but they cannot do so indefinitely without causing a complete collapse in access to care for their enrollees. Similarly, the transition to high-deductible plans will continue to attenuate health care cost growth until the transition is complete—a level versus a growth rate effect—at which point growth rates will revert to their long-run path.

What then is the long-run growth rate? Joseph Newhouse (1992) concluded that the long-term growth rate in health care spending is the consequence of technology growth—the introduction of new treatments and procedures and the diffusion of existing ones. Looking forward, we describe a variety of new technologies, some with modest health benefits but nearly all with large price tags, that could lead to billions of dollars in potential future health spending. For example, proton beam accelerators, which cost hundreds of millions of dollars to install yet offer no established advantages to patients over traditional treatment options, are expected to more than double in number between 2010 and 2014. New bioabsorbable stents are regarded as potentially energizing the otherwise moribund stent industry in the near future (Zamanian 2013). The near doubling of medical technology investment between 2009 and 2013 (Medmarket Diligence 2013) and stepped-up patent activity for medical devices, coupled with relatively strong performance in health care stocks, all suggest a reemergence of technology growth. Using these and other data, we predict that the real yearly growth rate of health spending will be 1.2 percent plus GDP growth over the next two decades. This estimate is bracketed by the Medicare Actuary Report (1.15 percent plus GDP growth through 2037; Boards of Trustees 2013) and the Congressional Budget Office (CBO) (1.3 percent plus GDP growth; CBO 2013).

We acknowledge that our prediction comes with a wide confidence interval. The impact of the ACA is not yet known, particularly the potential cost-saving effects of accountable care organizations (ACOs). Many knowledgeable observers already see the signs of a transformation in the U.S. health care system. As Len Nichols (2013, p. 7) interprets the situation:

. . . a good metaphor for the U.S. health care system today is the opening sweeping panorama [in The Sound of Music] followed by the crescendo of Julie Andrews’ voice singing “The Hills are Alive” with the sound of care process redesigns and incentive changes designed to make better outcomes sustainable at lower total cost.

If ACOs eschew building proton beam accelerators, change how patients pay for cost-ineffective technologies, and begin to slow the innovation cycle (Finkelstein 2007), we could see the promised transformation. Alternatively, the technology pipeline could start up again, bringing continued rapid growth in health care costs (perhaps coupled with sluggish GDP growth). Our prediction of 1.2 percent plus GDP growth, which implies that 23 percent of U.S. GDP will be spent on health care by 2032, is somewhere in between—perhaps not as uplifting as “The Hills Are Alive,” but not quite as scary as the cult horror film The Hills Have Eyes.

I. Facts about the Slowdown

What is the evidence on the slowdown in health spending? We begin with the macroeconomic evidence, adapted from Victor Fuchs (2013) and shown in figure 1. The graph shows smoothed inflation-adjusted growth rates in health care spending and demonstrates their sharp decline since roughly 2005, with a particularly dramatic drop in the 2010s. The decline is not quite so dramatic, however, in comparison to GDP growth. Until the most recent few years, health care spending growth tracked the downward spiral in GDP growth, and indeed, the share of health care spending to GDP has not yet dropped below 17 percent, still the highest in the world by far. Despite the considerable commentary about the recent slowdown in health care spending growth, average health spending growth between September 2012 and September 2013 exceeded GDP growth by 0.8 percent.

Figure 1. Total National Health Expendituresa and GDP Growth,b 1993–2013c.

a. Monthly health expenditure data are calculated by the Altarum Institute, and monthly GDP is estimated by Macroeconomic Advisors.

b. We use the implicit GDP deflator from Macroeconomic Advisors to deflate nominal national health expenditures.

c. Graph is smoothed using a density estimator with an Epanechnikov kernel and a bandwidth of 2 months. This graph is inspired by Fuchs (2013).

To illustrate the pitfalls associated with making forecasts about a slowdown in spending based on just a few years of data (Fuchs 2013), note that figure 1 shows a similar slowdown in the early 1990s, when the share of GDP spent on health care even declined slightly (from 13.7 percent in 1993 to 13.6 percent in 1994). The decline at the time was seen as a welcome correction (or even “revolution”) arising from greater competition in response to the growth of health maintenance organizations (HMOs) and the anticipation of the Clinton health care reforms. Robert G. Dederick, an economist, noted at the time that “the medical sector is not immune to what goes on in the economy. . . . It’s not as out of touch as many people seem to think.”³ Similarly, a Merrill Lynch vice president noted that “[p]hysicians are anticipating change and really beginning to change their practices.”⁴ Despite this enthusiasm, it is sobering to note that the drop-off was short-lived. By the late 1990s, growth in health care expenditures had exceeded the long-term average of GDP plus 2.4 percent, and in 2000 annual real per capita growth was 7 percent.

Health care expenditures relative to potential GDP are shown in the two panels of figure 2, with levels in the upper panel and growth rates in the lower panel. In both panels we use two measures of health care spending— total spending (which was used earlier in figure 1) and personal health consumption expenditures, which are national health expenditures less medical sector purchases of structures and equipment and expenditures for noncommercial medical research.⁵ In many respects, this latter measure is a better representation of actual health care consumption flows. Table 1 provides a breakdown of average annual growth rates for the different series: total and personal health care expenditures exhibit very similar growth rates, with the exception that total health care spending rose by less during the most recent recession than did personal health care spending.

Figure 2. Total National Health Spending and Personal Health Expendituresa as a Fraction of Potential GDP,b 1970–2012.

a. Total national and personal health expenditures are from the CMS National Health Expenditures Accounts from 1970–2011, with 2012 estimates provided by the Altarum Institute.

b. Nominal Potential GDP is estimated by the Congressional Budget Office. Yearly potential GDP is the average of quarterly estimates.

Table 1.

Average Yearly U.S. Health Spending Growth during Four Periods (Percent)

	1990–2001	2001–07	2007–09	2009–12
Real health spending
Total national health expenditures	4.6	4.8	2.9	2.5
Personal health care	4.6	4.5	3.6	2.5
Total national health expenditures as a share of potential GDP	1.4	2.0	0.6	0.8
Personal health care as a share of potential GDP	1.4	1.7	1.3	0.8
Per capita total national health expenditures	3.5	3.8	2.1	1.7
Per capita personal health care	3.5	3.5	2.8	1.7
Real per enrollee spending
Per enrollee Medicare national health expendituresa	4.0	5.3	3.4	1.2
Per enrollee Medicaid national health expendituresb	3.8	0.2	0.5	−1.4
Private premiumb	8.2	6.2	2.4	3.5

Source and methods: Health expenditures are from the CMS National Health Expenditures between 1990 and 2011, with 2012 estimates for national and personal expenditures provided by the Altarum Institute. Real values are deflated by the GDP deflator published by the Bureau of Economic Analysis, and potential GDP is published by the Congressional Budget Office. Private premiums are estimated by the Kaiser Family Foundation.

^aSeries only goes to 2011.

^bSeries starts in 1999.

The slowdown in health care spending is not unique to the United States. Figure 3, which uses aggregate health care spending from OECD data, demonstrates a slowdown between 2009 and 2011 that is more abrupt in other developed economies than in the United States. The sudden stop to spending growth in OECD countries likely reflected the necessity (and ability) of centrally financed health care insurance programs, such as the National Health Service in England, to hold the line on budgets by “cutting wages, reducing hospital staff and beds, and increasing copayments for patients.”⁶

Figure 3. Real Spending Growth in Total Health and Public Health, United States and Other OECD Countries, 2001–11.

a. U.S. total and personal spending per capita is from the CMS National Health Expenditures Accounts, and is deflated using the GDP deflator from the Bureau of Economic Analysis.

b. Spending growth for other OECD countries is calculated as a weighted average of real per capita PPP health spending using data from the Organization for Economic Cooperation and Development.

By contrast, Medicare is an entitlement program obligated to reimburse any bills submitted. There were a few modest attempts during the recent recession to scale back rates of Medicare reimbursement or restrict the services covered. State Medicaid administrators were better able to cut reimbursement rates and restrict eligibility over this time, but their budgets were strained by the sudden inflow of newly eligible (and often jobless) Medicaid enrollees. We return to these points below in section II.

Finally, we consider an alternative measure of growth in the health care sector: health care sector employment growth. Since 57 percent of overall health care expenditures are labor costs (Turner and Hughes-Cromwick 2013), it seems unlikely that we would expect to see a permanent bending of the cost curve without a commensurate shift in employment rates.

Figure 4 shows annual (smoothed) 12-month growth rates in the health care sector using two data surveys. The Current Population Survey (CPS) shows a strong pattern of growth in health care employment through the recent recession (2007–09), followed by a drop in 2010–11, then a jump back up in mid-2012, followed in turn by a drop in late 2012. (The CPS monthly estimates show much more variability than the smoothed estimates we show here.)

Figure 4. Year-over-Year Growth in Healtha Employment for the Current Population Survey (CPS) and the Current Employment Statistics (CES), by Month, 2001–13.

Source: Current Employment Statistics and author calculations based on the Current Population Survey Merged Outgoing Rotation Groups.

a. The health industry is defined according to NAICS codes 621, 622, and 623 from the 2002 version of the North American Industry Classification System.

b. Current employment statistics estimates are seasonally adjusted, and estimates of the number of employed in the CPS are smoothed using a 13-month-window moving average.

c. The fraction employed in the CES is the number of health employees over the total number of nonfarm workers.

By contrast, as measured by the Current Employment Statistics (CES) data, employment shows a remarkably constant growth rate in the health care sector since 1991, which has fluctuated around 2 percent per annum through business cycle contractions and expansions (figure 4).⁷

Why might the two measures of employment for the health care sector be so different? First, it should be noted that the two series are very similar in how they depict total employment growth (see appendix, figure A1). Second, there are some differences between the surveys in their treatment of multiple jobs held by the same worker and the CPS coverage of the self-employed, which are not reflected in the CES (Bowler and Morisi 2006). The third difference is in the surveys’ sample sizes; the CPS is derived from a sample of about 60,000 workers, compared to roughly 160,000 firms that cover 400,000 workplaces in the CES (Bowler and Morisi 2006). This size difference creates greater sampling variability in the CPS, particularly for sectoral-specific growth measures.⁸ And while even the CES has shown a recent slowdown in job growth, taken together these employment data suggest caution in predicting a permanent bending of the health care cost curve.

II. Factors That Might Have Reduced Cost Growth

In this section, we line up the usual suspects implicated in (or credited with) the decline in health care cost growth.

II. A. The Affordable Care Act of 2010

One explanation for the decline in health care spending growth, popular among Democrats, is the implementation (or anticipation of the implementation) of the ACA of 2010 (Unger 2013), with David Cutler (2013) concluding that “the ACA is a significant part of the reason” for the downturn,⁹ and the Council of Economic Advisors stating that “[t]he evidence . . . suggests that the ACA is already contributing to lower spending and price growth and that these effects will grow in the years ahead” (CEA 2013), p. 24).

The problem with this explanation is that thus far, the cost-saving effects of the ACA have been mixed. Some characteristics of the ACA increase costs, such as the extension of insurance coverage to dependents up to age 26. One ACO pilot reduced costs by less than 2 percent (Colla and others 2012), although a private Massachusetts initiative was able to save substantially more (McWilliams, Landon, and Chernew 2013; Song and others 2012). Most importantly, the cost-saving components of the ACA are not yet fully implemented, and they certainly could not explain why health care cost growth began to moderate in 2006, when Barack Obama was still a senator from Illinois.

However, two other features of the ACA may have made a difference a few years after the start of the slowdown. The first feature is rule changes in Medicare that were phased in starting in 2010, such as lower payments to Medicare Advantage plans and other providers (saving $17 billion over three years) as well as reduced payments to hospitals with poor quality measures (saving $230 to $280 million per year starting in 2012). These changes may have reduced national spending growth by 0.2 percentage points in the past three years (CEA 2013). Furthermore, these reductions in Medicare payment rates have spillover effects onto commercial plans— as Medicare rates change, commercial rates change in the same direction (Clemens and Gottlieb 2013; White 2013). While the magnitude of these effects is still debated, payment reductions in Medicare could have been mirrored by payment reductions in private plans, with current estimates suggesting dollar-for-dollar reduction spillovers.

The second feature of the ACA that may have made a difference is the anticipatory effects the law may have triggered among providers and insurers preparing for competition through the new health exchanges and alternative payments models. While this is possible—as the private Massachusetts initiative indicates—we do not see direct and widespread evidence of cost-saving strategies being adopted by providers, nor do we see evidence from Massachusetts that private health care costs are growing more slowly there than in the rest of the country.

II.B. The Great Recession

Others (primarily Republicans) have instead attributed the downturn in health care expenditures to the recession.¹⁰ Cutler and Sahni (2013) considered the role of the recession in explaining the downturn and estimated an elasticity of about one with respect to a 5-year average GDP. While their model predicts a rebound of health care spending as GDP picks up, there is still a roughly one-percentage-point drop in health care cost growth that cannot be explained by GDP growth per se.

We find that the time-series association between growth in GDP and growth in health care expenditures depends critically on the specified lag structure. In regression models using aggregate data on health care expenditures and GDP from 1970–2011, we find, like Cutler and Sahni (2013), a coefficient of about one using a 5-year geometric average. But the results were sensitive to the lag structure, with a 3-year geometric average yielding an estimate of 0.21 and a 7-year geometric average implying an estimate of 1.58. (appendix, table A3). Furthermore, a state-level time-series cross-section regression including year dummy variables yielded no significant relationship between state-level income changes and state-level health care spending.¹¹

One explanation for the wide range of results is that different components of U.S. health care spending respond quite differently to policy levers and other changes in the economy, some of which are likely correlated with GDP growth. For example, GDP growth generates tax revenue growth, which allows greater expansion of existing government-financed health care systems; the long-run elasticity of health care expenditures with regard to aggregate income is estimated to be roughly one (Getzen 1992); see Borger, Rutherford, and Won 2008).¹² The short-run fall in tax revenue during the Great Recession played a key role in reducing European health care spending (OECD, 2013), but as noted earlier, similar effects were muted in the United States; Medicare was largely insulated from budgetary cuts during the recession, while Medicaid spending actually increased as enrollment surged. The fraction of workers with private health insurance continued its gradual decline, but that trend had begun a decade earlier.

Some support for these hypotheses comes from aggregate time-series regressions that we carried out, considering each component of health care spending separately (see appendix, table A2). Using a 5-year lag, we demonstrate that Medicare expenditure growth appears unassociated with GDP growth, with a coefficient not significantly different from zero.¹³ By contrast, private health insurance is very strongly associated with GDP growth, with a coefficient of 3.0, suggesting that firms are more willing to insure their workers, and pay higher premiums, during economic upturns. Medicaid is somewhere in between, consistent with the pro-cyclical impact of GDP on state fiscal budgets moderated by the anti-cyclical effects of rising enrollment during recessions.

Another explanation for the downturn in health spending is that it has been driven by demand: enrollees are less likely to seek care because their income has declined. Yet most estimates of the income elasticity of demand are clustered near zero (e.g., McClellan and Skinner 2006; Borger, Rutherford, and Won 2008). Nor do Michael Levine and Melinda Buntin (2013) find that Medicare enrollees with financial downturns reduce their health care utilization by more than those who have not suffered such downturns. Recall, however, that these Medicare enrollees are largely insulated from the sting of copayments and deductibles. As we argue next, the landscape has changed dramatically in the United States for non-Medicare patients, who are exposed to much greater outof- pocket payments, making utilization far more sensitive to financial stress (Lusardi, Schneider, and Tufano 2010).

II. C. Patient Cost Sharing

One fundamental change that could explain the slowdown in health care costs is the rise in cost sharing and consequent rise in the price of services facing patients. For example, in a Kaiser/HRET survey of employer-sponsored health benefits, the fraction of employers offering a high-deductible plan grew from 4 percent in 2005 to 31 percent in 2011. Figure 5 shows the deductible facing the average employee with employer-provided health insurance with a deductible since 2006, using data from the same Kaiser/HRET survey. The dollar amounts reported on the vertical axis are the amounts the workers are potentially liable for, and not what they actually paid (which will reflect individual decisions made with regard to health). This figure shows that for people with an insurance plan that has a generalized annual deductible, the average size of the deductible more than doubled between 2005 and 2011.

Figure 5. Average Health Insurance Deductible for Individualsa across All Plan Types among Covered Workers with a Deductible, 2006–13.

Source: Kaiser HRET Survey of Employer-Sponsored Health Benefits, 2013, inflation-adjusted using the GDP deflator (through 2013)Q2).

a. Family plans have higher deductibles but similar growth patterns.

As a result of higher deductibles, out-of-pocket costs grew for patients enrolled in private insurance. Carolina-Nicole Herrera and others (2013) find that out-of-pocket payments grew at an average annual rate of 8 percent between 2007 and 2011; during this time, insurer spending on medical expenditures grew at an annual rate of 4.9 percent.¹⁴ Alexander Ryu and others (2013) find similar sharp increases in out-of-pocket spending for private insurance plans between 2007 and 2011; 61 percent for emergency room charges, 39 percent for brand-name drugs, 36 percent for hospital admissions, and 23 percent for outpatient visits. Based on the RAND health insurance study’s estimate of price elasticity of demand for health services of −0.2 (Manning and others 1987), an 8 percent per year nominal increase in out-of-pocket costs (which is approximately 6.4 percent real) would lead to an estimated decline in utilization of about 1.3 percent per year among patients of private plans.¹⁵

An additional factor leading to higher prices for working-age adults is the rising number of people who lacked insurance during this period. The percentage of the population ages 19–64 who were uninsured rose from 18 percent in 2005 to 20 percent in 2010 before dropping somewhat to 19 percent in 2012, owing to the legislated coverage of dependents up to age 26 under the ACA. For adults ages 26–49, uninsured rates continued their climb from 22 percent in 2010 to 24 percent in 2012 (Collins and others 2013).

Our discussion of the growing role of out-of pocket payments may appear to contrast with that of Katherine Baicker and Dana Goldman (2011), who note that as a share of total spending out-of-pocket payments have been falling over time. One reason for this apparent inconsistency is that there have been large increases in the use of post-acute services, such as skilled nursing facilities and home health care, that are covered by Medicare and Medicaid with very little cost sharing (Chandra, Dalton, and Holmes 2013). By contrast, as we noted earlier, rates of increase in out-of-pocket spending for private health insurance and the rise in the number of those uninsured have led to higher out-of-pocket prices facing households.

We hypothesize that this new landscape of high deductibles and copayments resulted in an interactive effect in which greater financial exposure amplified previously modest income effects. Further supporting this hypothesis are the findings of Annamaria Lusardi, Daniel Schneider, and Peter Tufano (2010), who conducted surveys in five countries during the depths of the Great Recession. On net, 19.5 percent of Americans responded in the survey that they reduced routine physician visits, compared to 6.6 percent of respondents in France and 3.6 percent in Germany, both countries with modest copayments. In Canada and Great Britain, countries with few if any copayments, there was no net decline in the rates of reported physician visits.¹⁶

II.D. Prices versus Quantities

Spending growth is the consequence of increasing prices or increasing quantities—or both. The distinction between prices and quantities is central to our explanation of the downturn, because Medicare, Medicaid, and commercial insurers manage prices and quantities so differently. Medicare performs little utilization review, and its only cost-control tool is to reduce reimbursement rates or, in a few cases, to restrict coverage. Medicaid manages reimbursements even more vigorously, for example by cutting reimbursements rates so much that enrollees essentially lack access to specialty care and newer technologies (Sack 2010).

Commercial payers, on the other hand, are much smaller than government payers and are far more likely to be affected by the rising power of the provider market.¹⁷ Rather than paying less to hospitals and physicians, as Medicare and Medicaid do, they attempt instead to slow health care cost growth on the demand side by increasing deductibles and copayments and by putting enrollees in tiered networks (where patients pay higher copayments to access high-cost hospitals).

Is the slowing growth of health care spending a consequence of falling prices or shrinking utilization? This simple question is surprisingly complex to answer. First, as Gerald Anderson and others (2003) have argued, the high level of U.S. health care spending reflects higher prices, although it is not known whether their argument applies to the growth of spending. Second, in the short run, price increases are likely to reflect the growth of unmeasured quantities; for example, if hospitals begin charging more to insert a stent, does that mean the price of the same stent went up, or that the hospitals are now using drug-eluting stents when in the past they used bare-metal stents? Third, a price index is quite difficult to measure in health care, since it does not capture improved survival rates and quality of life resulting from technology gains in health (Cutler and Newhouse 1998). Fourth, the well-known problems of Laspeyres price-indexes—that they utilized base-period weights and ignore substitution effects—is particularly pronounced in health care, where new therapies and new conditions can dramatically alter the distribution of spending across categories.

Of course, recognizing that there are serious problems with measurement does not mean that nothing can be learned. The problem of quantity and quality changes masquerading as price changes can be ameliorated by comparing relative price differences across payers—Medicare, Medicaid, and commercial. For example, prices paid by employer-sponsored health insurers for inpatient admissions grew at an annual rate of 5.2 percent between 2007 and 2011, while intensity only grew at an annual rate of 1.3 percent annually (Herrera and others 2013). However, during this period Medicare’s reimbursements to hospitals were just above inflation and its reimbursements for outpatient care were just below inflation.¹⁸ Thus we can conclude that, even in the presence of biases in measuring price increases for inpatient services, private (commercial) prices rose considerably more rapidly than Medicare prices.

The slowdown in overall Medicare spending could have also occurred because of the growth in managed care contracts—that is, a capitated rather than fee-for-service payment for what is now one-quarter of all Medicare enrollees. While it is difficult to know utilization trends in managed care (since these are not generally reported), Katherine Baicker, Michael Chernew, and Jacob Robbins (2013) estimate a pronounced spillover effect; health systems treating more managed care patients also treat their feefor- service patients more conservatively (also see Glied and Zivin 2002). Managed care could have also effected a reduction in the price per enrollee. Cutler and Sahni (2013) use internal data from Office of the Actuary of the Centers for Medicare and Medicaid Services (CMS) to estimate that during 2009–11, reimbursement rates declined by 1.8 percent annually, largely because of managed care reimbursement cuts. This means the actual utilization would not have fallen by as much as suggested by the decline in aggregate expenditures.¹⁹

Finally, there is substantial evidence from across states that during this period, Medicaid continued to cut provider fees and either implicitly or explicitly limit access to high-cost services, such as specialists (Sack 2010). Medicaid has always paid marginal cost or below, but during the Great Recession provider reimbursements have fallen further; the Medicaid-to- Medicare fee ratios for physicians declined from 72 percent to 66 percent between 2008 and 2011 (Zuckerman and Goin 2012).

In sum, our data suggest that price growth rates behaved very differently between commercial plans (the most rapid rate of growth), Medicare (little change in real terms), and Medicaid (some evidence of real price reductions). We turn next to the question of how innovation and the diffusion of new technologies have affected growth in the quantity of health care.

II.E. Technology Growth and Diffusion

The pace of technological growth is likely to have affected recent health care expenditure growth. A substantial percentage of the growth in health care costs since the 1980s has been the diffusion of new and expensive technologies; cardiac catheterization laboratories, hip and knee replacements, advanced intensive care unit (ICU) facilities, and new pharmaceutical approaches to treating cancer and other diseases. The diffusion of some of these technologies has indeed slowed. Drug spending actually declined in 2012, not only because of the rising share of generics but also because of a scarcity of new blockbuster drugs (Thomas 2013). But based on the technology pipeline, we believe that there will be a bounce-back in technology growth and innovation.

In earlier work, we found it useful to consider a typology of health care technologies to understand cost growth in health care, and we invoke the same classification here (Chandra and Skinner 2012). Category I treatments have high average productivity and are responsible for most of the increase in longevity. Category I treatments are either very low-cost, or high-cost but highly cost-effective, such as anti-retrovirals for HIV and AIDS. But even when they include high-cost treatments like anti-retrovirals, Category I technologies have not been a first-order driver of cost growth, and we do not expect them to have a discernible impact on overall expenditure growth in the years to come. Category II treatments have heterogeneous benefits across patients—with some patients gaining valuable health benefits but others receiving little incremental value. And Category III treatments—such as proton-beam therapy, or the aggressive use of ICU beds—both are expensive and have unknown (or no) incremental benefits.

Figure 6 shows per-enrollee Medicare fee-for-service rates of growth between 1994 and 2010 for a variety of treatments, with rates normalized to 1.0 in 1994. In the cases of cardiac stenting and bypass surgery, there was a dramatic run-up in their use during the 1990s for heart attacks and other heart disease, with a particularly rapid rise in the use of stents. During the mid-2000s, however, several randomized trials suggested that the use of stents for the most common types of heart disease offered only very modest benefits (e.g., Boden and others 2007), leading to a downturn in the use of these procedures. The downturn in the combined use of stents and coronary artery bypass grafts (CABG, or bypass surgery) is even more marked, as shown in figure 6. This change was not unique to cardiovascular procedures; as can be seen, the slow decline was observed for all inpatient surgical procedures as well.

Figure 6. Utilization Rates for Selected Procedures in the Medicare Fee-for-Service Population over Age 65, 1994–2009.

Source: Dartmouth Atlas of Health Care, various years.

In part, some of the decline in cardiac inpatient procedures could have been the consequence of a gradual shift during this period from inpatient to outpatient procedures—that is, if the operations were increasingly performed in outpatient surgical centers. Levine and Buntin (2013) correctly note that these types of technological innovations might be expected to reduce costs on a per-procedure basis, at least while the shift was taking place. At the same time, the introduction of new bioabsorbable stents—which are absorbed into the arterial walls after several months—has led some observers to forecast a rebound in market growth for stents (Zamanian 2013), potentially signaling a renewed rise in inpatient cardiac procedures and their corollary costs.

Other treatments, considered in more detail in table 2 using data from the Dartmouth Atlas of Health Care,²⁰ continued to grow in the later 2000s, but at a slower pace. For example, back surgery, a Category II treatment, grew at an annual rate of 6.5 percent during 1996–2001, before slowing to 2.5 percent growth in 2001–06 and 1.8 percent in 2006–10. And while a few Category III treatments, such as ICU days in the last six months of life (figure 6), continued to expand, table 2 shows clearly that many inpatient surgical procedures had turned the corner by 2006.

Table 2.

Annual Rates of Change of Per-Enrollee Utilization of Selected Surgical Procedures (Percent)

Procedure	1996–2001	2001–06	2006–10
All surgery	1.3	0.0	−3.4
Abdominal aortic aneurysm repair	1.9	−2.1	−3.4
Back surgery	6.5	2.5	1.8
Coronary angiography	3.8	−1.8	−7.2
Coronary artery bypass grafting (CABG)	−1.6	−6.6	−7.4
Percutaneous coronary interventions	8.0	2.4	−10.9
Carotid endarterectomy	−1.1	−5.6	−6.1
Cholecystectomy	−0.6	−3.6	−4.2
Hospitalization for hip fracture	−0.9	−1.6	−2.4
Hip replacement	3.7	2.4	4.0
Knee replacement	2.6	7.3	1.6
Mastectomy for breast cancer	−2.7	−8.7	−6.2
Resection for colon cancer	−0.3	−3.9	−16.2
Aortic/mitral valve replacement	2.7	0.1	0.2

Source: Dartmouth Atlas of Health Care.

Has the use of technology slowed down differentially between Medicare and commercial insurance? Some evidence for this comes from David Lee and Frank Levy (2012), who document the slowdown in Category II magnetic resonance imaging (MRIs) and X-ray computed tomography (CT scans) during the mid-2000s; for their sample of privately insured patients, growth was essentially flat, while for Medicare CT scans, growth continued at about 5 percent annually (Lee and Levy 2012). The anecdotal evidence from Lee and Levy (2012) lends support to the view that Medicare has less leverage than private insurance companies in slowing the diffusion of Category II technologies.

II.F. Other Potential Explanations

A variety of other factors might have influenced health care cost growth. Levine and Buntin (2013) find that younger and healthier Medicare enrollees, the leading edge of Baby Boomers, influenced expenditure growth, but the effects they find are small. Wages and compensation for health care employees could have fallen, but we do not find evidence for this using the wage data in the CPS (appendix, figure A2). Finally, as noted above, overall health care employment has continued to climb in the face of declining hospital admissions and physician visits, making it unlikely that health care providers could pass along savings through lower prices.

III. Accounting for Recent Growth in Costs

The major drivers of the health care slowdown are changes in relative prices to consumers and providers (having both income and substitution effects) and technological growth—each of which will affect private, Medicare, and Medicaid patients differently. Figure 7 shows rates of per enrollee utilization for each of these different components of health care expenditures. Despite the recent slowdown in Medicare spending growth, total Medicare spending has risen the most rapidly, even during the recent downturn, due to growth in both spending and enrollment (although growth in real terms per beneficiary slowed to 0.4 percent in 2012; Kronick and Po 2013). The overall contribution of private insurance expenditures to health care cost growth has been moderating since the early 2000s because of a decline in the share of the population covered. Even though average real premiums per enrollee have continued to rise by about 2.7 percent per year, most of the increase stemmed from increased prices, and utilization has not grown by much (HCCI 2012).

Figure 7. Real Health Spending per Enrollee and per Capita, by Payer, 2001–11.

Sources: Total health spending per capita and per enrollee expenditures are from the CMS National Health Expenditures Accounts from 2001–2011, with 2012 estimates for national expenditures provided by the Altarum Institute. Private premiums are for a representative individual with no dependents, and are estimated by the Kaiser Family Foundation. Real values are deflated by the GDP deflator published by the Bureau of Economic Analysis.

Table 3 provides our best-guess benchmark 2007–11 measures of the components of growth in Medicare, Medicaid, and private insurance. These include price growth per enrollee, quantity (or intensity) growth per enrollee, and growth in enrollment rates. While there are other components of health care spending, such as public health, we focus on Medicare, Medicaid, and private commercial insurance payments, whose shares are 30, 22, and 48 percent of this subset of spending, respectively.²¹

Table 3.

Growth Rates of Prices, Utilization, and Enrollments by Payer, 2007–11 (Percent)

Payer	Price growth (real)	Utilization growth	Enrollment growth	Share of payments	Share of enrollees
Medicarea	0.9	1.4	2.5	30	16
Medicaidb	−0.4	−0.1	4.7	22	19
Privatec	2.7	0.7	−1.3	48	65
Total (weighted)	1.5	0.7	0.4	100	100

Notes: The GDP deflator was 1.6 percent and was subtracted from column 1. To estimate aggregate price growth, we take a weighted average of component growth rates for the years 2007–2011, where the weights are total spending by component-year.

We take per enrollee spending growth, enrollment growth, share of payments, and share of enrollees from the data sets listed in figure 7. Utilization growth (g_u) is calculated from nominal spending growth (g_n) and price growth (g_p) according to the equation:

$$1+g_u=\frac{1+g_n}{1+g_p}.$$

Total price and utilization growth rates are weighted by the share of payments. Enrollment growth is weighted by the share of enrollees.

^aPrice growth in Medicare is based on estimates of nominal price growth of 2.5 percent per year between 2007 and 2010 by Levine and Buntin (2013).

^bMedicaid price growth is taken from Zuckerman and Goin (2012), who estimate that physician fees have risen by 4.9 percent in Medicaid between 2008 and 2012, which is an annualized growth rate of 1.2 percent.

^cPrice growth for private payers is calculated based on Health Care Cost Institute calculations of price and utilization growth for inpatient, outpatient, physician, and prescription drug benefits in their 2010, 2011, and 2012 cost growth reports.

In table 3, we calculate growth in total spending and enrollment using the CMS National Health Expenditures data, supplemented by data on private premiums from the Kaiser Family Foundation. Price growth in Medicare, Medicaid, and private insurance is taken from a literature review, and then we calculate utilization growth to be the residual “real” growth rate.²² Considering first the Medicare market, we use the Levine and Buntin (2013) estimates of changes in Medicare price growth.²³ Medicaid price growth varies widely by state, but we abstract from this regional variation by taking the aggregate estimates of price growth from Stephen Zuckerman and Dana Goin (2012). Finally, our estimates of real price growth are based on the HCCI (2012) estimates for 2007–11.²⁴

We can use this table to consider several straightforward hypotheses. First, how much did the ability of state governments to hold the line on additional Medicaid expenditures contribute to the slowdown? Under the counterfactual that Medicaid prices would have risen as rapidly as private health insurance prices, the answer is 0.5 percent slower growth during this period, which is half of the unexplained 1-percentage-point drop estimated by Cutler and Sahni (2013).

A second question is, How much have utilization controls at private providers (chiefly through increased cost sharing) and in Medicaid (chiefly through the restriction of certain types of care such as specialist visits) contributed to the slowdown? This is a more difficult question to answer, because Medicaid enrollees have been becoming healthier over time. Between 2007 and 2010, because of the weak economy, enrollment in Medicaid rose far faster among healthier adults than among those who were aged or disabled. Based on a simple back-of-the-envelope calculation, average utilization between 2007 and 2010 would be expected to fall by about 0.8 percent per year simply because the average age of Medicaid patients is dropping. This means age-adjusted utilization growth in Medicaid is closer to 0.7 percent per year, which is similar to utilization growth in private spending. After this correction, if we were to assume that utilization growth for both private plans and Medicaid were equal to the 1.4 percent utilization growth in Medicare, then overall spending growth would have been 0.4 percent higher.

IV. Will Cost Growth Revert to Long-Run Trends?

Two decades ago, Joseph Newhouse addressed the question of why health care continued to rise so much faster than GDP growth (Newhouse 1992). After ruling out a number of alternative explanations, he arrived at the root cause: technological innovations. Other factors, such as changes in insurance generosity and coverage, had increased utilization, but these were one-time shifts rather than a movement in the long-term growth rate. Thus our next task is to attempt to disentangle what might be the short-run effects that might be expected to moderate or disappear, in order to discern the underlying long-run growth rate in health care expenditures, with a special focus on technology growth and diffusion. We recognize the risks of prognostication. As John Kenneth Galbraith noted, “There are two types of forecasters; those who don’t know and those who don’t know they don’t know.”²⁵ Thus these predictions come with very wide confidence intervals.

We consider three complementary approaches to assessing the prospect for long-term growth. The first approach peers into the technology pipeline to see whether it really has slowed, and also whether industry observers are optimistic with regard to future growth in devices and surgical procedures. The second approach considers relative movements in health sector stock prices. The third and final approach is to use the estimates from table 3 above to develop our best estimate of the long-run trends separate from short-run or transitory effects.

IV.A. The Technology Pipeline

Let us consider, qualitatively, developments in health care technology. We first note, as have Cutler and Sahni (2013), a distinct lack of new blockbuster drugs; combined with a rising share of generic drug sales, the pharmaceutical industry has exhibited slow growth in recent years. Yet drugs account for roughly one-tenth of total health care spending.

The story is different, however, for devices and new surgical procedures, particularly in cardiovascular care. For example, there has been rapid diffusion in left ventricular assistance devices (LVADs). These devices were originally used to keep potential heart transplant patients alive, but now they are increasingly used for long-term therapy among patients with heart failure. LVADs are very expensive, costing over $300,000 in the first year (Rogers and others 2012). While the market now is relatively modest (and includes former Vice President Richard Cheney), the sector’s growth rate is projected to be 10 to 15 percent annually.²⁶ More worrisome with regard to costs is the interest among clinicians in expanding the use of LVADs to patients over age 70, as well as to less seriously ill (but far more plentiful) patients, such as those with Class III heart failure (Stewart and Stevenson 2011).

Another cardiovascular care procedure being developed is the new transaortic valve replacement (TAVR). While many elderly people have valve disorders, for most the risk of open surgery is not worth the benefit of replacing the valve. These valve replacements, by contrast, do not require open heart surgery, but are instead inserted using a catheter threaded through the femoral artery, a much less invasive procedure. There is consequently a much larger potential market for the use of TAVRs among the thousands of frail elderly people suffering from poor circulation. Additional new technologies with robust growth projections include left atrial appendage closure for atrial fibrillation patients, and bioabsorbable stents, designed to be reabsorbed into the body after several months and thus reducing the risk of adverse long-term complications.

There are also many new developments in the treatment of cancers. One that has been gaining considerable attention from both policymakers and investors is the increased use of proton beam therapy for prostate cancer patients. While there is no evidence that outcomes arising from this treatment are better than alternative treatments, such as radiation therapy or prostatectomy (the removal of the prostate), it costs roughly double what they cost—$50,000 per course of treatment as compared with $25,000 or even less. This is an example of a Category III treatment, expensive but with no proven value. The willingness of Medicare (and hence private insurance) to pay at least the average total cost of this treatment creates a strong incentive to invest in the large fixed costs of the proton beam facility (hundreds of millions of dollars) and an equally strong incentive to run through as many prostate cancer patients as possible to pay off the bonds.

Figure 8 shows the number of actual and expected proton beam facilities in the United States.²⁷ After a slow start in the 1990s, there has been a rapid acceleration in the planning of these facilities, often with two hospitals in the same region each planning their own facility (Gold 2013). The total number of proton beam accelerators (planned and built) is expected to double between 2010 and 2014 alone, suggesting that a major driver of costs—Category III technologies—is still being discovered.

Figure 8. Number of Proton Beam Accelerator Facilities in the U.S., 1990–2016a.

Source: Particle Therapy Co-Operative Group.

a. Includes all facilities operating, under construction, or planned as of 2013.

Predicting aggregate trends in technology development, rather than just specific cases, is much harder. One proxy for future growth is new medical technology funding, which has nearly doubled between 2009 and 2013 (MedMarket Diligence 2013). Another proxy for future technology growth is the number of patents approved. The number of U.S. patents approved for medical device manufacturing remained steady at about 9,000 per year between 2000 and 2004, fell to about 7,000 per year between 2005 and 2010, and then rose to record levels of about 13,000 per year between 2011 and 2012.²⁸ As noted by Levine and Buntin (2013), the number of approved patents is partly determined by internal policies of the U.S. Patent Office, so it is not a good proxy for year-on-year investment research and development. However, what the recent surge in approved patents does suggest is that the pipeline is full of technologies that may appear on the market in the coming two decades.

Why is the United States particularly vulnerable to technology growth that is highly costly yet yields benefits that are often small or not reliably measured? One reason may be that Medicare is legislated to pay for any treatments that will not actually cause harm. Until recently, private insurance companies have found it difficult to refuse payment for treatments already approved by Medicare, so the United States is a particularly fertile environment for such growth (Chandra and Skinner 2012). To return to the example of the proton beam accelerator, consider that Germany, Italy, Spain, the United Kingdom, and France combined have about the same population as the United States, but as of 2013 they have among them just 10 proton beam accelerators (and an average of 1.2 per year coming online in the near future), as compared to 15 accelerators in the United States (and an average of 3 per year coming online).

IV.B. Financial Markets

Another approach to projecting long-term growth in the health care sector is to examine the response of financial markets to both the overall slowdown in health care expenditures, and more specifically to the introduction of the ACA. Mohamad Al-Ississ and Nolan Miller (2013) use Massachusetts Republican Scott Brown’s election to the U.S. Senate as an instrument to measure the impact of a decline in the probability of the ACA’s passage.²⁹ Based on the 2.1 percent abnormal returns to the stock market in response to the fall in the probability of passage, they conclude that the ACA was viewed by Wall Street as supporting cost-containment—although the expansion of Medicaid was anticipated to have a beneficial impact on hospital stocks.

But what about the health care cost growth slowdown more generally—can we detect a general downturn in health sector stocks since the mid-2000s? The comparison is complicated by the stock market collapse during the recession, but it is still instructive to compare the long-term relative trends in equity returns for the health sector on the New York Stock Exchange (figure 9). The evidence from these returns is not consistent with the hypothesis that new efforts to contain costs (whether through the ACA, employers, or insurance companies) have at last put a brake on the development and diffusion of profitable Category II and Category III treatments.³⁰

Figure 9. New York Stock Exchange Cumulative Returns, Health Sector and NYSE Composite Index, 2003–13.

Source: New York Stock Exchange Index Services.

a. The NYSE currently includes 109 companies that are classified in the health care sector according to the Industry Classification Benchmark, which is proprietary to FTSE International Limited and Dow Jones & Company, Inc. The most common company types in the index are pharmaceutical companies, health care providers, and medical equipment companies.

IV.C. Predicting Cost Growth

To make our prediction, we begin with the 2007–11 growth data in table 3, and we consider how these different factors each might be expected to be transformed, given the evidence discussed in the earlier sections. Consider first the private insurance market. Prices might continue to rise faster than inflation, as they have since the mid-2000s. However, we usually assume that when prices rise because of market power, such as hospital consolidation (as in Gaynor and Town 2012) or lack of price transparency, at some point the limits of consolidation and opaqueness will be reached, limiting the scope by which prices may be expected to grow forever. Similarly, the introduction of price-reducing reforms such as bundled payments, reference pricing, or an improvement in price transparency would be expected to slow price growth down substantially in the shorter term, but would have little effect on long-term growth in prices.³¹ We conjecture that real price increases in the private insurance sector will drop from the 2.7 percent annual growth in 2007–11 to 1.5 percent over the next two decades.

In the longer term, it is unlikely that the growth rate in Medicaid payments can continue to fall so far behind Medicare payments while still maintaining a network of providers willing to accept Medicaid patients. Thus we assume that both Medicaid and Medicare payments will continue to rise by a rate of 0.9 percent plus inflation (the current Medicare rate of growth in reimbursement, from table 3). This would be substantially slower cost growth than in private plans because of Medicare’s ability to impose administratively set prices.

What then will happen to real quantity increases in health care? As noted in table 3, real Medicare utilization per enrollee rose at 1.4 percent annually based on 2007–11 growth patterns, which we assume will continue. ³² This is a conservative estimate, since there are reasons (as noted above) to believe that a new wave of innovations could increase Medicare spending. Private plans continue to have avenues for restricting spending, such as defined-contribution plans, limited networks, value-based-insurance design, and transparency tools. But the rollout of new high-deductible plans is nearly finished, and so we expect utilization growth to bounce back to 1.4 percent per year. We further assume a long-term growth in per capita Medicaid utilization of 1 percent, which is marginally higher than the age-adjusted real growth in utilization for Medicaid during 2007–11.

Finally, what about enrollment growth? The ACA will increase the fraction of uninsured who qualify for Medicaid, but the best estimate of how that will affect spending is roughly $1,000 per individual (Finkelstein and others 2012), so even if 3 percent of the population becomes newly eligible for Medicaid, this will still only boost per capita health care expenditures by $30. Of course, aging baby boomers will tend to increase enrollment in the Medicare program (which grew at 2.4 percent between 2008 and 2012), but we account for such changes by applying a general aging index, which actuaries have estimated to be roughly 0.4 percent annually, using recent data (Yamamoto 2013).

Adding these three effects (1.2 percent for price growth plus 1.3 percent for utilization growth and 0.4 percent for aging) yields an annual real growth in health care spending of 2.9 percent. Longer-term real GDP growth projections based on the Congressional Budget Office and the U.S. Census Bureau suggest real GDP per capita growth of 1.7 percent from 2012 through 2032.³³ In sum, we end up at GDP plus 1.2 percent; certainly below the historical record of GDP plus 2.4 percent (Fuchs 2013) and not very different from current (2013) data (figure 1), but still consistent with a long-term growth rate in the health sector from 17.9 percent currently to roughly 23 percent in 2032.

V. Conclusion

There has been considerable media and government attention to the question of whether health care costs have moderated. In this paper, we reconsider the existing evidence on this slowdown by taking a more disaggregated view of the health care sector. We first note that while all measures of health care spending and utilization point toward a recent slowdown, it was not the first time this has happened; health care costs as a fraction of GDP had actually declined in the early 1990s before resuming their strong upward trend in the later 1990s. And certainly the remarkable stability in one measure of health care employment growth reinforces Victor Fuchs’ (2013) caution against inferring too much about the next two decades from just a few years of data.

Second, we present evidence that aggregate health care spending growth might not exhibit a stable association with GDP growth over the medium term. Unlike other types of consumption, health care in the United States is an aggregate of very different systems—private, Medicare, and Medicaid— whose dynamic paths of quantity and costs do not move in lockstep with one another, particularly during a recession or business cycle expansion. For this reason, we argue that researchers should consider a more disaggregated model of health care, complete with a specification of the factors such as reimbursement rates, prices, and technological developments for public and private services.

Third, in explaining the downturn in health care spending, we have placed a greater emphasis than previous researchers on the rise in cost sharing in the private insurance market. Although we consider these changes in a static framework—one can’t continue to raise co-payment rates forever— there may also be dynamic effects arising from these changes. According to Amy Finkelstein (2007) and Jeffrey Clemens (2011), the rise of high-deductible health insurance will both deliver a short-run reduction in the quantity of services demanded and also exert a longer-term impact on the incentive to adopt and pay for new innovations. That said, we do not see evidence of such a long-term impact on innovation, at least based on the relative growth of health sector stock prices.

Fourth, we predict continued long-term growth in real per capita health care spending that will exceed GDP. Our best guess is GDP plus 1.2 percent, which puts us in the middle of the pack between Medicare actuaries and the CBO, albeit with wide confidence intervals. Even this modest estimate is not a cause for celebration; Charles Roehrig (2012) has called attention to the “triangle of painful choices,” which outlines the set of very unpleasant options facing the United States even in the face of a “moderate” GDP plus 1 percent growth rate in health care costs. The pain includes some combination of increases in tax rates or drastic cuts in non-health spending, and does not capture the additional unpleasantness of private health insurance premium hikes soaking up any real wage growth for the median worker (Auerbach and Kellermann 2011).

Finally, and more optimistically, we also recognize that the structure and balance of power among providers and insurers may be undergoing fundamental changes. For example, private insurers, emboldened by an increase in market share as they receive more patients from health exchanges and the Medicare Advantage program, may begin to push back against the coverage of Category III treatments.³⁴ Nascent signs of this are apparent in the isolated decisions by some private insurers to no longer cover proton-beam therapy.³⁵

Similarly, the adoption of ACOs in Medicare and the move toward bundled payments could encourage providers to switch from expensive and unproven therapies to cheaper ones. Many of these initiatives involve private partnerships with leading integrated delivery systems, such as Intermountain Healthcare’s collaboration with General Electric. Moreover, on both sides of the political aisle there is a consensus that fee-for-service creates incentives for overuse. And while the exact solutions to this problem may differ on the spectrum of market versus regulatory approaches to technology management, a Republican-led Congress would continue the move toward payment reform. Yet ultimately, all policy solutions must also address the concern that the long-term technology pipeline will continue to deliver new technology with large price tags but very modest health benefits.

Figure 10. Difference between Yearly Growth in National Health Expenditure (NHE) and GDP, 2007–13.

Sources: Monthly Health Expenditure Data are calculated by the Altarum Institute, and monthly GDP is estimated by Macroeconomic Advisors.

Note: Graph is smoothed using a kernel density estimator with an Epanechnikov kernel and a bandwidth of 2 months.

Appendix

Figure A1 Year-over-Year Employment Growth for the Current Population Survey (CPS) and the Current Employment Statistics (CES), by Month, 2001–13

Source: Current Employment Statistics and author calculations based on the Current Population Survey Merged Outgoing Rotation Groups, which are maintained by the NBER. Current employment statistics measure the growth in all non-farm jobs, and the Current Population Survey sample is restricted to those aged 18–64.

Figure A2. Real Median Income of Full-Time Employees (All vs. Health Sector), 1990–2012

Source: Authors’ calculations based on the Current Population Survey Merged Outgoing Rotation Groups, which are maintained by the NBER. Sample includes employees ages 18 to 64 and excludes self-employed workers, and nominal wages are deflated using the GDP deflator published by the Bureau of Economic Analysis.

Table A1.

Growth in Real National Health Expenditures (NHE)^a and Real GDP,^b by Year, 2001–13 (Percent)

Year	NHE	GDP	NHE minus GDP
2001–02	8.8	2.7	6.1
2002–03	6.1	2.5	3.6
2003–04	4.0	3.8	0.2
2004–05	3.5	3.3	0.2
2005–06	3.2	2.0	1.2
2006–07	3.6	1.9	1.7
2007–08	2.4	1.0	1.4
2008–09	3.7	−4.3	8.0
2009–10	2.6	3.3	−0.7
2010–11	1.2	1.5	−0.3
2011–12	3.5	3.6	−0.1
2012–13	2.5	1.2	1.3
Average	3.8	1.9	1.9
Average, 2006–13	2.8	1.2	1.6

^aEstimates of nominal national health expenditures each July by year are from the Altarum Institute. We use the implicit GDP deflator from Macroeconomic Advisors to deflate nominal national health expenditures.

^bEstimates of GDP each July are from Macroeconomic Advisors.

Table A2.

Real Per Capita Spending Growth by Payer vs. GDP Growth Rate, 1970–2012 (Dependent variable: growth in real per capita costs paid by)

	Total	Personal	Medicare	Medicaid	Private Insurance	Out-of- pocket spending
Panel A: no lags
Real, per capita GDP growth (Current period)	0.165 (0.105)	0.0625 (0.0939)	−0.217 (0.256)	−0.370 (0.433)	0.512** (0.233)	0.508*** (0.168)
R2 (Model 1)	0.044	0.007	0.017	0.022	0.089	0.164
Panel B: with 3 lags
Sum of current and lagged real per capita GDP growth	0.553***	0.544***	−0.463	0.313	0.980*	0.962**
Wald F statistic	6.77	8.81	.998	.323	3.7	7.33
P-value	0.013	0.005	0.324	0.573	0.062	0.010
R2	0.189	0.170	0.025	0.105	0.190	0.229
Panel C: with 5 lags
Sum of current and lagged real per capita GDP growth	1.398***	1.187***	−0.820	2.179**	3.076***	1.589***
Wald F statistic	54.4	41.4	1.06	6.35	58.4	15
P-value	0.000	0.000	0.311	0.016	0.000	0.000
R2	0.561	0.437	0.040	0.244	0.570	0.338
First year	1970	1970	1970	1970	1970	1970
Last year	2012	2012	2011	2011	2011	2011
N	43	43	42	42	42	42

Source: Total national health expenditures are from the CMS National Health Expenditures Accounts from 1970–2011, with 2012 estimates for total and personal spending provided by the Altarum Institute. The gross domestic product and GDP deflator are published by the Bureau of Economic Analysis.

^*p < .10,

^**p < .05,

^***p < .01

Note: Standard errors in parentheses. Results are from OLS regressions with robust standard errors. F statistics are against the null that the current and all lagged coefficients are all equal to zero.

Table A3.

Health Spending Growth vs. GDP Growth Rate

	Dependent variable: growth in real per capita total national health expenditures
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Real per capita	0.165
GDP growth	(0.105)
Average 2-year		0.164
GDP growth		(0.116)
Average 3-year			0.183
GDP growth			(0.167)
Average 4-year				0.428**
GDP growth				(0.196)
Average 5-year					0.938***
GDP growth					(0.192)
Average 6-year						1.328***
GDP growth						(0.176)
Average 7-year							1.475***
GDP growth							(0.207)
Average 8-year								1.488***
GDP growth								(0.249)
Constant	0.0385***	0.0385***	0.0381***	0.0337***	0.0240***	0.0161***	0.0125***	0.0116**
	(0.00320)	(0.00351)	(0.00442)	(0.00448)	(0.00391)	(0.00359)	(0.00439)	(0.00545)
First year	1970	1970	1970	1970	1970	1970	1970	1970
Last year	2012	2012	2012	2012	2012	2012	2012	2012
R2	0.0439	0.0277	0.0225	0.0827	0.270	0.402	0.388	0.327
N	43	43	43	43	43	43	43	43

Source: Total national health expenditures are from the CMS National Health Expenditure Accounts from 1970–2011, with 2012 estimates provided by the Altarum Institute. The gross domestic product and GDP deflator are published by the Bureau of Economic Analysis.

^*p < .10,

^**p < .05,

^***p < .01

Standard errors in parentheses. Results are from OLS regressions with robust standard errors.